Measuring the goodness of lineups: parameter estimation, question effects, and limits to the mock witness paradigm

Abstract

Lineups and photospreads can be biased against a criminal suspect and there is a need to measure this bias. The mock witness method has been accepted by eyewitness scientists since the 1970s as the paradigm for generating the data on which various metrics of bias are based. We note the reasons that structural lineup bias can lead to mistaken identification and we discuss the statistical metrics that are proposed to measure the degree of bias. Using mock witness data, we show that estimates of bias can vary as a function of the question asked of the mock witnesses. In addition, we note various lineup biases that are not measured with the mock witness paradigm, such as procedural biases and problems with propitious heterogeneity. Copyright © 1999 John Wiley & Sons, Ltd.